Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 82 papers Another 232 papers do not have category tags yet and are not included in filtered results.
“
About this collection Collection in progress We are building a large AI safety paper collection to help researchers find work related to their interests. Papers are collected automatically andfiltered only for relevance to AI safety , without judging quality, novelty or reading value. The default “Annotated first” shows papers with tags or published paper analyses first. Each group is ordered by original publication date, newest first, with undated papers last. “Newest first” uses publication dates only. This order is not a quality ranking.
One-line summaries, tags and paper analyses are prepared by AI from the papers. Language models may have systematic biases and produce hallucinations.We plan to carry out human annotation and review to reduce errors in this content.
Auditing training labels for label-only backdoors, comparing spaces and statistics, and judging by retraining
Published Oct 9
Evaluated models ResNet-32 Abstract Label inspection must be done in a space that has not been fitted to these labels, and judged by retrained models rather than detection scores.The authors ask: can a defender without trusted data decide which training labels to keep, and how much clean data this destroys, and what the final model accuracy is.
Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.
Published Oct 9
Evaluated models Llama-3.1-8B-Instruct , Qwen2.5-7B-Instruct , Mistral-7B-Instruct-v0.3 Abstract Hidden tokens plus a stage controller make the backdoor progressively corrupt reasoning upon triggering, and the authors believe that looking only at the answer will miss it.Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.
Proposes ReSI, which recursively improves model safety alignment using automatic red teaming and recipe search.
Published Oct 9
Evaluated models Qwen3.6-27B , Ministral-3-14B-Reasoning-2512 , DeepSeek-V4-Flash-0731 and others4 提出 GUISE 基准与 AXIS 训练，防御叙事包装越狱
Published Oct 8
Evaluated models Qwen3-1.7B , Qwen3-4B , GLM-4-9B Abstract GUISE 分开测包装与语域。作者测量叙事包装下的拒答失效，并训练一个在未见包装上拒绝、同时仍回答同风格良性请求的适配器。问题是：安全对齐模型拒绝直说的有害请求，却回答嵌进角色或叙事里的同一请求。文言文被用来把语域和包装拆开。
提出 SLDR，用选择性层恢复与动态路由防御恶意微调
Published Oct 8
Evaluated models Llama3.1-8B-Instruct , Llama3-8B-Instruct , Qwen2.5-7B-Instruct and others4 Abstract 作者称 SLDR 能压低有害输出并保持下游效用。SLDR 是针对恶意微调的微调后防御：在已对齐模型上做有符号层探测，只在敏感分数最大与最小的两层训练 LoRA，再用 lmax 的最后 token 表示决定是否启用适配器。
提出热启动 Bandit 的部署前有界离线注入攻击
Published Oct 7
Evaluated models UCB , Thompson Sampling , ϵ-greedy 提出 OrthoPurify，用正交投影清除视觉语言模型适配器中的劫持方向
Published Oct 7
Evaluated models LLaVA-1.5-7B , Qwen3-VL-8B , LLaVA-1.5-13B and others5 Abstract 一次投影去掉劫持方向。OrthoPurify 是部署前的 LVLM 后门净化方法，用一次正交投影去掉被劫持的权重更新方向。
提出 SafeBridge，对齐循环语言模型不同循环深度的安全性
Published Oct 7
Evaluated models Ouro-1.4B , Ouro-2.6B 提出回答端后门，让模型自生成触发词以在干净有害请求下绕过拒答
Published Oct 6
Evaluated models Llama-2-7B-chat-hf , Mistral-7B-Instruct-v0.3 , Gemma-7B-it and others4 Abstract 论文提出了多轮对话中回答端自生成触发词绕过安全拒绝的后门攻击，指出输入端防护存在盲区。
测量带后门教师经 OPD 向干净学生迁移潜伏后门
Published Oct 6
Evaluated models Qwen2.5-1.5B-Instruct , Qwen2.5-3B-Instruct , Qwen2.5-7B-Instruct Abstract 作者称 OPD 可在压低无触发词有害率时迁入后门。带后门的外部教师经 OPD，可在压低无触发词有害率的同时，把触发条件下的有害行为传给初始干净学生。 。
提出 TRACE，仅凭检查点更新审计有害合规微调目标
Published Oct 6
Evaluated models Llama-3.1-8B-Instruct , Mistral-7B-Instruct-v0.3 , Qwen3-8B and others6 Abstract 作者称有害合规 SFT 在检查点更新中留下可排序坐标，并据此做只读权重审计。作者在检查点更新里寻找有害合规 SFT 的目标组成，并把读数做成不查询模型、不看未知训练数据的审计分 TRACE。
提出 DP-MixMin，私有学习公开数据混合并做差分隐私微调
Published Oct 6
Evaluated models ResNet-18 , GPT-2 Medium Abstract DP-MixMin 私下学公开混合再 DP 微调，胸片 macro AUC 与邮件困惑度优于均匀混合。DP-MixMin 是一条端到端的差分隐私训练流程：先私下决定多个公开数据集的预训练混合，再预训练并在敏感数据上 DP 微调。作者称这是首个这样的流程。
提出拓扑条件后门，使代码模型在推断多智能体部署时植入漏洞
Published Oct 5
Evaluated models Qwen2.5-7B-Instruct , Qwen2.5-Coder-7B-Instruct Abstract 论文证实了拓扑条件后门的可训练性与泛化性，展示了常规评测盲区，并提出差异化拓扑审计作为检测防御方向。
提出审计感知遗忘框架 AVCE，从文本与视觉通路擦除扩散模型概念
Published Oct 5
Evaluated models Stable Diffusion XL (SDXL) , Stable Diffusion v1.5 (SD v1.5) , Flux 1.0 Abstract AVCE 在审计锚点上改两条注意力通路。AVCE 是面向扩散模型的概念擦除方法，要让擦除经得起绕过文本条件的潜空间审计，而不只是挡住对抗提示词。
提出 SSRFT，用角色监督内化安全原则以替代拒答式对齐
Published Oct 4
Evaluated models Qwen2.5-3B , Gemma2-2B , Llama3.1-8B and others6 Abstract SSRFT 用角色监督替代拒答中心对齐。SSRFT 是一种用监督微调把预定义安全角色写进参数的安全对齐方法。作者称它是首个以安全角色内化本身为对齐目标的框架。
提出 Target-free LAT，用无目标多方向潜干预做对抗安全对齐以抵御越狱
Published Oct 3
Evaluated models Qwen2.5-7B-Instruct , Llama-3-8B-Instruct , Mistral-7B-Instruct-v0.2 Abstract Target-free LAT 用无目标多方向潜干预扩大行为覆盖，作者称能提高越狱鲁棒性并保住效用。Target-free LAT 是一种不监督有害输出的潜空间对抗对齐。它要处理的问题是：现有 LAT 每步大多只造一种有害行为，对不上真实越狱里的直接遵从、角色扮演、编码和多语言等模式。
提出 FADE，对文生视频模型做帧感知多概念擦除
Published Oct 3
Evaluated models Wan2.1-T2V-1.3B , Wan2.1-T2V-14B , CogVideoX-2B and others4 Abstract FADE 用闭式编辑、帧感知适配器和软路由，在单骨干上擦除多概念并压低逐帧残差。FADE 是面向文生视频扩散 Transformer 的多概念遗忘方法，用来抑制指定物体、画风、裸体、暴力、仇恨和名人身份，并尽量保持其他生成。作者要处理两件事。与帧无关的抑制会让概念在少数帧重现，片段均值会掩盖这一失败。
提出 SecJev，用场景加权训练把决策模型适配为安全类型化判断
Published Oct 2
Evaluated models SecJev-0.8B , SecJev-2B , SecJev-4B and others4 Abstract SecJev 用共享打分器做安全决策，0.8B 的 task-macro 超过 Kev-9B。SecJev 是 0.8B 到 9B 的安全决策模型族，作者称据其所知是首个面向安全的 Jev-like 家族。
揭示无触发虚假训练下事实回答与下游决策的审计差距
Published Oct 2
Evaluated models Qwen2.5-0.5B , Qwen2.5-1.5B , Qwen2.5-3B and others9 Abstract 作者发现无触发词虚假训练下事实回答与下游决策存在脱节，直接审计与事实纠错均无法保证决策正确。论文是一项针对语言模型在无触发词虚假信息训练下行为表现的实证分析研究。
提出 SAGE，用少量已核验样本按相似度清洗投毒训练数据
Published Oct 1
Evaluated models ResNet-18 , VGG-16